Run Vector Search Queries

ANN Search

For Approximate Nearest Neighbors (ANN) search, MongoDB Vector Search finds vector embeddings in your data that are closest to the vector embedding in your query based on their proximity in multi-dimensional space and based on the number of neighbors that it considers. It uses the Hierarchical Navigable Small Worlds algorithm and finds the vector embeddings most similar to the vector embedding in your query without scanning every vector. Therefore, ANN search is ideal for querying large datasets without significant filtering.

ENN Search

For an Exact Nearest Neighbors (ENN) search, MongoDB Vector Search exhaustively searches all the indexed vector embeddings by calculating the distance between all the embeddings and finds the exact nearest neighbor for the vector embedding in your query. This is computationally intensive and might negatively impact query latency. Therefore, we recommend ENN searches for the following use-cases:

• You want to determine the recall and accuracy of your ANN query using the ideal, exact results for the ENN query.

• You want to query less than 10000 documents without having to tune the number of nearest neighbors to consider.

• Your want to include selective pre-filters in your query against collections where less than 5% of your data meets the given pre-filter.

If you enable automatic quantization, MongoDB Vector Search uses only the full-fidelity vectors for ENN queries.

Limitations

$vectorSearch can't be used in view definition and the following pipeline stages:

• $lookup sub-pipeline [1]

• $facet pipeline stage

MongoDB Vector Search Indexing

You must index the fields to search using the $vectorSearch stage inside a vectorSearch type index definition. You can index the following types of fields in a MongoDB Vector Search vectorSearch type index definition:

• Fields that contain vector embeddings as vector type.

• Additional fields as the filter type to enable vector search on pre-filtered data.

To learn more about these MongoDB Vector Search field types, see How to Index Fields for Vector Search.

MongoDB Vector Search Scoring

MongoDB Vector Search assigns a score, in a fixed range from 0 to 1 (where 0 indicates low similarity and 1 indicates high similarity), to every document that it returns.

The score is calculated according to the similarity function that you specify in the MongoDB Vector Search index definition. To learn more about the similarity options you can choose from, see About the Similarity Functions.

Each returned document includes the score as metadata. To return each document's score along with the result set, use a $project stage in your aggregation pipeline and configure the score as a field to project. In the score field, specify a $meta expression with the value vectorSearchScore. The syntax is as follows:

1
db.<collection>.aggregate([
2
  {
3
    "$vectorSearch": {
4
      <query-syntax>
5
    }
6
  },
7
  {
8
    "$project": {
9
      "<field-to-include>": 1,
10
      "<field-to-exclude>": 0,
11
      "score": { "$meta": "vectorSearchScore" }
12
    }
13
  }
14
])

MongoDB Vector Search Pre-Filtering

The $vectorSearch filter option matches BSON boolean, date, objectId, numeric, string, and UUID values, including arrays of these types.

You must index the fields that you want to filter your data by as the filter type in a vectorSearch type index definition. Filtering your data is useful to narrow the scope of your semantic search and ensure that not all vectors are considered for comparison.

MongoDB Vector Search supports the $vectorSearch filter option for the following MQL operators:

Parallel Query Execution Across Segments

We recommend dedicated Search Nodes to isolate vector search query processing. You might see improved query performance on the dedicated Search Nodes. Note that the high-CPU systems might provide more performance improvement. When MongoDB Vector Search runs on search nodes, MongoDB Vector Search parallelizes query execution across segments of data.

Parallelization of query processing improves the response time in many cases, such as queries on large datasets. Using intra-query parallelism during MongoDB Vector Search query processing utilizes more resources, but improves latency for each individual query.

Prerequisites

Before you run these examples, perform the following actions:

• Add the sample collection to your Atlas cluster.

• Create MongoDB Vector Search indexes for the collection. For instructions, see the Create a MongoDB Vector Search Index procedure and copy the configurations for the basic or filter examples in your desired language.

Basic ANN Example

The following query uses the $vectorSearch stage to search the plot_embedding_voyage_3_large field using vector embeddings for the string time travel. It considers up to 150 nearest neighbors, and returns 10 documents in the results. The query also specifies a $project stage to do the following:

• Exclude the _id field and include only the plot and title fields in the results.

• Add a field named score that shows the vector search score for each document in the results.

Filtered ANN Example

The following query filters the documents for movies released between January 01, 1955 and January 01, 1975 before performing the semantic search against the sample vector data. It uses the $and operator to perform a logical AND operation of the specified dates. It then searches the plot_embedding_voyage_3_large field in the filtered documents for 150 nearest neighbors using the vector embeddings for the string kids adventure, and returns 10 documents in the results. The query also specifies a $project stage to do the following:

• Exclude the _id field and include only plot, title, and year fields in the results.

• Add a field named score that shows the vector search score of the documents in the results.

MongoDB Vector Search filters the documents based on the year field value that ranges between 1955 and 1975. It returns documents that summarize children's adventures in the plot for movies released between 1955 and 1975.

ENN Example

The following query uses the $vectorSearch stage to search the plot_embedding_voyage_3_large field using vector embeddings for the string world war. It requests exact matches and limits the results to 10 documents only. The query also specifies a $project stage to do the following:

• Exclude the _id field and include only the plot, title, and year fields in the results.

• Add a field named score that shows the vector search score of the documents in the results.